I'm Siqi Fan (范嗣祺), a researcher at Institute for AI Industry Research, Tsinghua University (AIR, THU). Previously, I obtained my M.S. degree from the Institute of Automation, Chinese Academy of Sciences (CASIA) in 2022 and my B.E. degree from Shanghai Jiao Tong University (SJTU) in 2019.
My research interests lie broadly in Representation Learning in Complex Systems, spanning from the macro physical world to the micro biological world. With the goal of increasing human "Available Time" through AI technology, I primarily concentrate on
I love music and visual arts.
") does not match the recommended repository name for your site ("
").
", so that your site can be accessed directly at "http://
".
However, if the current repository name is intended, you can ignore this message by removing "{% include widgets/debug_repo_name.html %}
" in index.html
.
",
which does not match the baseurl
("
") configured in _config.yml
.
baseurl
in _config.yml
to "
".
IEEE (TIP, TCSVT, TVT, TIV, ITSM), IET (CV, CSR), PR, Neurocomputing
CVPR, ICCV, ECCV, ICRA, IROS, ITSC
Traffic scenes understanding and simulation testing @ ITSC'22
My exploration of vehicle-side environment perception began with drivable area detection (ITSC’20). Since then, I have proposed a series of perception algorithms, including an RGB 2D object detection approach designed for complex traffic environments (FII-CenterNet, T-VT’21), a semi-supervised learning approach for RGB 2D segmentation (CPCL, T-IP’22), an RGB-T segmentation approach tailored for challenging lighting conditions (SpiderMesh, TechReport’23), and a 3D segmentation approach for large-scale point clouds (SCF-Net, CVPR’21).
Compared to the well-studied vehicle-side perception, roadside perception faces several unique challenges, and the calibration noise caused by inevitable natural factors is one of them. Addressing that, a calibration-free BEV representation network is proposed to alleviate inaccurate calibration parameter problem (CBR, IROS’23). The development of roaside perception has been hindered by a lack of available data. On the one hand, a semantic-geometry decoupled contrastive learning framework is introduced to enhance roadside perception performance by leveraging vehicle-side data (IROAM, ICRA’25); On the other hand, the first large-scale real-world dataset for roadside cooperative perception is released, complete with benchmarks, to stimulate research in practical I2I perception (RCooper, CVPR’24).
Cooperative perception can significantly enhance individual perception performance by providing additional viewpoints and expanding the sensing field. A scene-level feature cooperative perception approach is proposed (EMIFF, ICRA’24). To enable interpretable, instance-level, and flexible feature interactions, the concept of query cooperation is introduced, along with a cooperative perception framework that allows query streams to flow among agents (QUEST, ICRA’24). Additionally, motion forecasting can also benefit from learning cooperative trajectory representations (NeurIPS’24). Beyond focusing on improving individual modules, a pioneering end-to-end cooperative autonomous driving framework is introduced (UniV2X, AAAI’25).
The microscopic biological system is intriguing but challenging. An all-atom framework is explored to enable consistent representation and interaction modeling across different biomolecules (PharMolixFM, TechReport’25).
Recent advances in LLMs have illuminated the path toward developing knowledgeable and versatile AI research assistants across various scientific domains. Multimodal large language models are particularly promising, as they bridge the semantic gap between natural language and other modalities such as molecules, proteins, and visual information. In this context, a multimodal large language model is proposed to assist biomedical research (BioMedGPT, J-BHI’24). Additionally, the task of optical chemical structure understanding is introduced and explored to facilitate molecule-centric scientific discovery (OCSU, TechReport’25).
Multi-agent cooperation holds great potential for solving complex scientific research tasks in an autonomous manner. To facilitate further exploration in this area, an agent platform specifically designed for biomedicine and life science is presented and open-sourced (OpenBioMed).